Research map / Label-efficient and robust learning / Recognition and long-tail
Deep Cross-species Feature Learning for Animal Face Recognition via Residual Interspecies Equivariant Network
Xiao Shi, Chenxue Yang, Xue Xia, Xiujuan Chai
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add residual inter-species feature equivariant module to fine-tuning (ablation) | Fine-tuning with ArcFace · Pig face verification | mean accuracy (%): 90.81 -> 92.83 | Improved |
| add residual inter-species feature equivariant module to fine-tuning (ablation) | Fine-tuning with ArcFace · Horse face verification | mean accuracy (%): 81.79 -> 82.24 | Improved |
| add animal facial feature fusion of lower half face (ablation) | RiseNet equivariant module · Pig face verification | mean accuracy (%): 92.83 -> 93.76 | Improved |
| add animal facial feature fusion of lower half face (ablation) | RiseNet equivariant module · Horse face verification | mean accuracy (%): 82.24 -> 82.56 | Improved |
| add dimension-reduction weight W (k*r x k) in fusion (ablation) | RiseNet · Pig face verification | mean accuracy (%): 93.27 -> 93.76 | Improved |
| lower fusion rate from 1 to 0.1 (ablation) | RiseNet without W · Pig face verification | mean accuracy (%): 92.74 -> 93.27 | Improved |
| lower fusion rate from 0.5 to 0.1 (ablation) | RiseNet without W · Pig face verification | mean accuracy (%): 92.82 -> 93.27 | Improved |
| lower fusion rate from 0.1 to 0.01 (ablation) | RiseNet without W · Pig face verification | mean accuracy (%): 93.27 -> 93.02 | Hurt |
| replace ArcFace trained from scratch with fine-tuning from human-face pretraining | ArcFace · Pig face verification | mean accuracy (%): 87.74 -> 90.81 | Improved |
| replace fine-tuning with RiseNet | Fine-tuning with ArcFace · Pig face verification | mean accuracy (%): 90.81 -> 93.76 | Improved |
| replace fine-tuning with RiseNet | Fine-tuning with ArcFace · Horse face verification | mean accuracy (%): 81.79 -> 82.56 | Improved |
| use domain adaptation DAN instead of fine-tuning on human faces | ArcFace · Pig face verification | mean accuracy (%): 89.72 vs 90.81 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- ArcFace — ArcFace: Additive Angular Margin Loss for Deep Face Recognition (CVPR 2019) · better on 6 of 6 shared results
- SphereFace — SphereFace: Deep Hypersphere Embedding for Face Recognition (CVPR 2017) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- AnimalWeb: A Large-Scale Hierarchical Dataset of Annotated Animal Faces (CVPR 2020 · similarity 0.89)
- Dense Interspecies Face Embedding (NeurIPS 2022 · similarity 0.87)
- Interspecies Knowledge Transfer for Facial Keypoint Detection (CVPR 2017 · similarity 0.87)
- PetFace: A Large-Scale Dataset and Benchmark for Animal Identification (ECCV 2024 · similarity 0.85)
- Cross-Domain Adaptation for Animal Pose Estimation (ICCV 2019 · similarity 0.85)
- SapiensID: Foundation for Human Recognition (CVPR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
